A Teleoperated Grasping Framework Based on Myoelectric Control and Stiffness Recognition

The dexterous operation of the bionic hand is challenging and involves multi-finger coordinated control, contact perception and contact force control. In traditional teleoperated robotic systems, teleoperating a bionic hand to grasp objects safely and stably remains a challenging task, especially for soft objects. To better adapt to grasping tasks involving hard and soft objects, a novel bionic hand grasping framework is proposed for teleoperation scenarios, achieving secure and stable manipulation through stiffness perception and grasping force control. Specifically, a multi-task learning method is proposed for simultaneous gesture and force level recognition form surface electromyography (sEMG), aiming to coordinately control the grasping pattern and force of the bionic hand. Furthermore, an optimized K-nearest neighbors (KNN) classification algorithm is proposed for object stiffness recognition based on three-dimensional (3D) contact force signals. Experiments demonstrate that the proposed framework exhibits superior performance in soft object grasping compared to sEMG-based bionic hand pattern control.